Emergency Alert System

August 15, 2024 · View on GitHub

Emergency Alert System

Features

Image Analysis: Analyzes images to detect emergencies.
Emergency Classification: Classifies emergencies into categories like fallen person, fire, medical emergency, etc.
Integration with MQTT: Publishes emergency states to an MQTT broker.

Integrations

graph TD;
    A[Collect images from cameras] --> C[Submit to OpenAI for analysis]
    C --> D[Publish results to Mosquitto]
    D --> E[Store data points in InfluxDB]
    E --> F[Alerting via Grafana Cloud]
    F --> G0[Grafana OnCall app]
    F --> G1[any phone number]
    F --> G2[instant messaging, email, webhooks, etc.]

https://motion-project.github.io/ collects images from cameras
this script, originally created just to display the images, has evolved to submit them to OpenAI for analysis
then publish the results to https://mosquitto.org/
https://github.com/koo5/iot2/ stores the data points in InfluxDB
https://grafana.com/products/cloud/ alerting is used to ping my phone when an emergency is detected.

MQTT

https://github.com/emqx/blog/blob/main/en/202111/popular-online-public-mqtt-brokers.md

Grafana Cloud

Grafana Cloud is super-complex, we need to add support for some easier alerting services, like Twilio, Telegram, which will not probably even integrate through MQTT anyway.

Todo

systemd unit template, for one unit for each camera.
add a switch to disable the lame GUI and run headless

i have zero idea about local models. As on OpenAI... i've burned through $0.22 for my testing so far .. so i'm imagining, one might run let's say 5 cameras, but first there is motion where you configure what constitutes an event, and then you add some delays and sleeps .. so i'm thinking some hundreds of images per day, maybe going into a couple dollars a day, but i imagine things will get significantly cheaper over time?

Requirements

Python 3.x
OpenAI API (optional, but works great)
Roboflow API (optional, experimental, disused)
Paho MQTT client
Fire (CLI)
Pygame (GUI)
mpv (GUI)
espeak (AUI :-) )

Installation

Setup venv:

virtualenv -p /usr/bin/python3.10 venv; . venv/bin/activate;

Install the required Python packages:
```
pip install -r requirements.txt
```

Usage

Run the main script with the desired options:

 MQTT_PORT=1883 MQTT_HOST=broker.hivemq.com OPENAI_API_KEY=`cat ~/secrets/OPENAI_API_KEY` ./src/main.py /var/lib/motion --lookback=5 --speak=True --CHATGPT=True